THE USE OF DERIVATIVE DYNAMIC TIME WARPING IN ANT COLONY INSPIRED CLUSTERING

The ECG data can be processed using feature extraction which strongly depends on data preprocessing used. This paper presents an application of swarm-based metaheuristic method inspired by the behavior of real ants in the nature - Ant Colony inspired clustering, a stochastic cooperative metaheuristics. The method is evaluated in the process of electrocardiogram interpretation and processing. It takes advantage of dynamic time warping method modification which produces satisfactory results comparable to feature extraction (which determines classical clinical parameters, such as amplitude and duration of important waves). Paper also considers the relevant steps to speed up the algorithm and make it robust in respect to the data used. Paper also discusses the use of standard clustering methods in the ECG processing using the dynamic time warping algorithm and provides a comparison with ant colony inspired clustering method. For pedagogic purposes, a framework for ant-colony related clustering methods has been created. The framework uses a VTK toolkit for 3D visualization. The dynamic time warping method is also compared to the automatic feature extraction method. Over the experiments, both methods yield quite similar results, thus the hybrid use of such distance measure is also considered.